Feat: Support pipeline-related configuration at the document level. (#17212)

### Summary

Feat: Support pipeline-related configuration at the document level.

web/src/components/builtin-pipeline-form-field.tsx

@@ -0,0 +1,67 @@
+import { SelectWithSearch } from '@/components/originui/select-with-search';
+import {
+  FormControl,
+  FormField,
+  FormItem,
+  FormLabel,
+  FormMessage,
+} from '@/components/ui/form';
+import { useFetchBuiltinPipelines } from '@/hooks/use-agent-request';
+import { cn } from '@/lib/utils';
+import { useFormContext } from 'react-hook-form';
+import { useTranslation } from 'react-i18next';
+
+export function BuiltinPipelineItem({
+  line = 2,
+  name = 'parser_id',
+}: {
+  line?: 1 | 2;
+  name?: string;
+}) {
+  const { t } = useTranslation();
+  const form = useFormContext();
+  const { options: builtinPipelineOptions } = useFetchBuiltinPipelines();
+
+  return (
+    <FormField
+      control={form.control}
+      name={name}
+      render={({ field }) => (
+        <FormItem className="items-center space-y-0">
+          <div
+            className={cn('flex', {
+              'items-center': line === 1,
+              'flex-col gap-1': line === 2,
+            })}
+          >
+            <FormLabel
+              required
+              className={cn('text-sm whitespace-wrap', {
+                'w-1/4': line === 1,
+              })}
+            >
+              {t('knowledgeConfiguration.builtIn')}
+            </FormLabel>
+            <div
+              className={cn('text-muted-foreground', { 'w-3/4': line === 1 })}
+            >
+              <FormControl>
+                <SelectWithSearch
+                  {...field}
+                  placeholder={t(
+                    'knowledgeConfiguration.chunkMethodPlaceholder',
+                  )}
+                  options={builtinPipelineOptions}
+                />
+              </FormControl>
+            </div>
+          </div>
+          <div className="flex pt-1">
+            <div className={line === 1 ? 'w-1/4' : ''}></div>
+            <FormMessage />
+          </div>
+        </FormItem>
+      )}
+    />
+  );
+}

web/src/components/chunk-method-dialog/index.tsx

@@ -24,7 +24,6 @@ import {
   ChunkMethodItem,
   EnableTocToggle,
   ImageContextWindow,
-  ParseTypeItem,
 } from '@/pages/dataset/dataset-setting/configuration/common-item';
 import { zodResolver } from '@hookform/resolvers/zod';
 import omit from 'lodash/omit';
@@ -44,6 +43,7 @@ import { ExcelToHtmlFormField } from '../excel-to-html-form-field';
 import { LayoutRecognizeFormField } from '../layout-recognize-form-field';
 import { MaxTokenNumberFormField } from '../max-token-number-from-field';
 import { MinerUOptionsFormField } from '../mineru-options-form-field';
+import { ParseTypeItem } from '../parse-type-form-field';
 import { ButtonLoading } from '../ui/button';
 import { Input } from '../ui/input';
 import { DynamicPageRange } from './dynamic-page-range';

web/src/components/document-pipeline-dialog/index.tsx

@@ -0,0 +1,109 @@
+import { BuiltinPipelineItem } from '@/components/builtin-pipeline-form-field';
+import { DataFlowSelect } from '@/components/data-pipeline-select';
+import { ParseTypeItem } from '@/components/parse-type-form-field';
+import PipelineOperatorTabs from '@/components/pipeline-operator-tabs';
+import { ButtonLoading } from '@/components/ui/button';
+import {
+  Dialog,
+  DialogContent,
+  DialogFooter,
+  DialogHeader,
+  DialogTitle,
+} from '@/components/ui/dialog';
+import { Form } from '@/components/ui/form';
+import { ParseType } from '@/constants/knowledge';
+import { IModalProps } from '@/interfaces/common';
+import { IChangeParserRequestBody } from '@/interfaces/request/document';
+import { useCallback } from 'react';
+import { useTranslation } from 'react-i18next';
+import {
+  IDocumentPipelineDialogProps,
+  useDocumentPipelineForm,
+} from './use-document-pipeline-form';
+
+const FormId = 'DocumentPipelineDialogForm';
+
+interface IProps
+  extends IModalProps<IChangeParserRequestBody>, IDocumentPipelineDialogProps {
+  loading: boolean;
+}
+
+export function DocumentPipelineDialog({
+  hideModal,
+  onOk,
+  parserId,
+  pipelineId,
+  parserConfig,
+  loading,
+}: IProps) {
+  const { t } = useTranslation();
+
+  const {
+    form,
+    parseType,
+    operatorNodes,
+    operatorNodesLoading,
+    activeTab,
+    setActiveTab,
+    handleOperatorValuesChange,
+    showOperatorTabs,
+    buildSubmitData,
+  } = useDocumentPipelineForm({ parserId, pipelineId, parserConfig });
+
+  const onSubmit = useCallback(
+    async (data: Parameters<typeof buildSubmitData>[0]) => {
+      const ret = await onOk?.(buildSubmitData(data));
+      if (ret) {
+        hideModal?.();
+      }
+    },
+    [buildSubmitData, hideModal, onOk],
+  );
+
+  return (
+    <Dialog open onOpenChange={hideModal}>
+      <DialogContent className="max-w-[50vw] text-text-primary">
+        <DialogHeader>
+          <DialogTitle>{t('knowledgeDetails.chunkMethod')}</DialogTitle>
+        </DialogHeader>
+
+        <Form {...form}>
+          <form
+            onSubmit={form.handleSubmit(onSubmit)}
+            className="space-y-6 max-h-[70vh] overflow-auto -mx-6 px-10 py-5"
+            id={FormId}
+          >
+            <ParseTypeItem />
+            {parseType === ParseType.BuiltIn && <BuiltinPipelineItem />}
+            {parseType === ParseType.Pipeline && (
+              <DataFlowSelect
+                isMult={false}
+                showToDataPipeline={true}
+                formFieldName="pipeline_id"
+              />
+            )}
+            {showOperatorTabs && (
+              <PipelineOperatorTabs
+                nodes={operatorNodes}
+                value={activeTab}
+                onValueChange={setActiveTab}
+                onOperatorValuesChange={handleOperatorValuesChange}
+              />
+            )}
+          </form>
+        </Form>
+        <DialogFooter>
+          <ButtonLoading
+            type="submit"
+            form={FormId}
+            loading={
+              loading || (operatorNodesLoading && operatorNodes.length === 0)
+            }
+          >
+            {t('common.save')}
+          </ButtonLoading>
+        </DialogFooter>
+      </DialogContent>
+    </Dialog>
+  );
+}

web/src/components/document-pipeline-dialog/use-document-pipeline-form.ts

@@ -0,0 +1,204 @@
+import { ParseType } from '@/constants/knowledge';
+import {
+  useActiveTab,
+  usePipelineOperatorNodes,
+  useResetParserConfigOnPipelineChange,
+} from '@/hooks/use-pipeline-operator';
+import { IChangeParserRequestBody } from '@/interfaces/request/document';
+import {
+  getOperatorType,
+  transformApiConfigToForm,
+  transformFormConfigToApi,
+} from '@/utils/pipeline-operator';
+import { zodResolver } from '@hookform/resolvers/zod';
+import { useCallback, useEffect, useMemo } from 'react';
+import { useForm, useWatch } from 'react-hook-form';
+import { useTranslation } from 'react-i18next';
+import { z } from 'zod';
+
+export interface IDocumentPipelineDialogProps {
+  parserId: string;
+  pipelineId?: string;
+  parserConfig?: Record<string, any>;
+}
+
+/**
+ * Converts the saved parser_config (API format, keyed by operator id) to the
+ * form format expected by the operator tabs, mirroring
+ * useFetchDatasetSettingOnMount on the dataset setting page.
+ */
+function transformSavedParserConfigToForm(
+  parserConfig?: Record<string, any>,
+): Record<string, any> {
+  if (
+    !parserConfig ||
+    typeof parserConfig !== 'object' ||
+    Array.isArray(parserConfig)
+  ) {
+    return {};
+  }
+
+  const hasPipelineKeys = Object.keys(parserConfig).some((key) =>
+    key.includes(':'),
+  );
+  if (!hasPipelineKeys) {
+    return parserConfig;
+  }
+
+  const formParserConfig: Record<string, any> = {};
+  for (const [operatorId, config] of Object.entries(parserConfig)) {
+    formParserConfig[operatorId] = transformApiConfigToForm(
+      getOperatorType(operatorId),
+      config as Record<string, any>,
+    );
+  }
+  return formParserConfig;
+}
+
+export function useDocumentPipelineForm({
+  parserId,
+  pipelineId,
+  parserConfig,
+}: IDocumentPipelineDialogProps) {
+  const { t } = useTranslation();
+
+  const FormSchema = useMemo(
+    () =>
+      z
+        .object({
+          parseType: z.nativeEnum(ParseType),
+          parser_id: z.string(),
+          pipeline_id: z.string().optional(),
+          parser_config: z.record(z.string(), z.any()).optional(),
+        })
+        .superRefine((data, ctx) => {
+          if (data.parseType === ParseType.BuiltIn && !data.parser_id.trim()) {
+            ctx.addIssue({
+              path: ['parser_id'],
+              message: t('common.pleaseSelect'),
+              code: 'custom',
+            });
+          }
+          if (data.parseType === ParseType.Pipeline && !data.pipeline_id) {
+            ctx.addIssue({
+              path: ['pipeline_id'],
+              message: t('common.pleaseSelect'),
+              code: 'custom',
+            });
+          }
+        }),
+    [t],
+  );
+
+  const form = useForm<z.infer<typeof FormSchema>>({
+    resolver: zodResolver(FormSchema),
+    defaultValues: {
+      parseType: pipelineId ? ParseType.Pipeline : ParseType.BuiltIn,
+      parser_id: parserId || '',
+      pipeline_id: pipelineId || '',
+      parser_config: transformSavedParserConfigToForm(parserConfig),
+    },
+  });
+
+  const parseType = useWatch({
+    control: form.control,
+    name: 'parseType',
+  });
+  const selectedDataFlowId = useWatch({
+    control: form.control,
+    name: 'pipeline_id',
+  });
+  const selectedBuiltinId = useWatch({
+    control: form.control,
+    name: 'parser_id',
+  });
+
+  const isPipelineMode = parseType === ParseType.Pipeline;
+  const selectedPipelineId = isPipelineMode
+    ? selectedDataFlowId
+    : selectedBuiltinId;
+  // The saved parser_config only belongs to the pipeline (and parse type) it
+  // was saved with — expose its id only while that exact pipeline is selected,
+  // so that after switching, defaults come purely from the new pipeline's DSL.
+  const savedParseType = pipelineId ? ParseType.Pipeline : ParseType.BuiltIn;
+  const savedPipelineId =
+    parseType === savedParseType
+      ? isPipelineMode
+        ? pipelineId
+        : parserId
+      : undefined;
+
+  const { operatorNodes, loading: operatorNodesLoading } =
+    usePipelineOperatorNodes(
+      selectedPipelineId,
+      selectedPipelineId && selectedPipelineId === savedPipelineId
+        ? parserConfig
+        : undefined,
+      !isPipelineMode,
+    );
+
+  useResetParserConfigOnPipelineChange(
+    form,
+    selectedPipelineId,
+    savedPipelineId,
+    operatorNodes,
+  );
+
+  const { activeTab, setActiveTab } = useActiveTab(operatorNodes);
+
+  useEffect(() => {
+    if (parseType === ParseType.BuiltIn) {
+      form.setValue('pipeline_id', '');
+    }
+  }, [parseType, form]);
+
+  const handleOperatorValuesChange = useCallback(
+    (operatorId: string, values: any) => {
+      const currentParserConfig = form.getValues('parser_config') || {};
+      form.setValue('parser_config', {
+        ...currentParserConfig,
+        [operatorId]: values,
+      });
+    },
+    [form],
+  );
+
+  const showOperatorTabs =
+    operatorNodes.length > 0 &&
+    ((parseType === ParseType.Pipeline && !!selectedDataFlowId) ||
+      (parseType === ParseType.BuiltIn && !!selectedBuiltinId));
+
+  const buildSubmitData = useCallback(
+    (data: z.infer<typeof FormSchema>): IChangeParserRequestBody => {
+      const transformedConfig: Record<string, any> = {};
+      for (const [operatorId, config] of Object.entries(
+        data.parser_config ?? {},
+      )) {
+        transformedConfig[operatorId] = transformFormConfigToApi(
+          getOperatorType(operatorId),
+          config as Record<string, any>,
+        );
+      }
+
+      const isPipeline = data.parseType === ParseType.Pipeline;
+      return {
+        parser_id: isPipeline ? '' : data.parser_id,
+        pipeline_id: isPipeline ? data.pipeline_id : '',
+        parser_config: transformedConfig,
+      };
+    },
+    [],
+  );
+
+  return {
+    form,
+    parseType,
+    operatorNodes,
+    operatorNodesLoading,
+    activeTab,
+    setActiveTab,
+    handleOperatorValuesChange,
+    showOperatorTabs,
+    buildSubmitData,
+  };
+}

web/src/components/parse-type-form-field.tsx

@@ -0,0 +1,73 @@
+import {
+  FormControl,
+  FormField,
+  FormItem,
+  FormLabel,
+  FormMessage,
+} from '@/components/ui/form';
+import { Radio } from '@/components/ui/radio';
+import { ParseType } from '@/constants/knowledge';
+import { cn } from '@/lib/utils';
+import { useFormContext } from 'react-hook-form';
+import { useTranslation } from 'react-i18next';
+
+export function ParseTypeItem({
+  line = 2,
+  name = 'parseType',
+}: {
+  line?: number;
+  name?: string;
+}) {
+  const { t } = useTranslation();
+  const form = useFormContext();
+
+  return (
+    <FormField
+      control={form.control}
+      name={name}
+      render={({ field }) => (
+        <FormItem className="items-center space-y-0">
+          <div
+            className={cn('flex', {
+              'items-center': line === 1,
+              'flex-col gap-1': line === 2,
+            })}
+          >
+            <FormLabel
+              className={cn('text-sm  whitespace-wrap ', {
+                'w-1/4': line === 1,
+              })}
+            >
+              {t('knowledgeConfiguration.parseType')}
+            </FormLabel>
+            <div
+              className={cn('text-muted-foreground', { 'w-3/4': line === 1 })}
+            >
+              <FormControl>
+                <Radio.Group {...field}>
+                  <div
+                    className={cn(
+                      'flex gap-2 justify-between text-muted-foreground',
+                      line === 1 ? 'w-1/2' : 'w-3/4',
+                    )}
+                  >
+                    <Radio value={ParseType.BuiltIn}>
+                      {t('knowledgeConfiguration.builtIn')}
+                    </Radio>
+                    <Radio value={ParseType.Pipeline}>
+                      {t('knowledgeConfiguration.manualSetup')}
+                    </Radio>
+                  </div>
+                </Radio.Group>
+              </FormControl>
+            </div>
+          </div>
+          <div className="flex pt-1">
+            <div className={line === 1 ? 'w-1/4' : ''}></div>
+            <FormMessage />
+          </div>
+        </FormItem>
+      )}
+    />
+  );
+}

web/src/components/pipeline-operator-tabs/pipeline-operator-form.tsx

@@ -1,12 +1,12 @@
 import { Operator } from '@/constants/agent';
 import { RAGFlowNodeType } from '@/interfaces/database/agent';
+import ExtractorForm from '@/pages/agent/form/extractor-form';
+import ParserForm from '@/pages/agent/form/parser-form';
+import TitleChunkerForm from '@/pages/agent/form/title-chunker-form';
+import TokenChunkerForm from '@/pages/agent/form/token-chunker-form';
+import TokenizerForm from '@/pages/agent/form/tokenizer-form';
+import { getOperatorType } from '@/utils/pipeline-operator';
 import { memo, useCallback } from 'react';
-import ExtractorForm from '../../agent/form/extractor-form';
-import ParserForm from '../../agent/form/parser-form';
-import TitleChunkerForm from '../../agent/form/title-chunker-form';
-import TokenChunkerForm from '../../agent/form/token-chunker-form';
-import TokenizerForm from '../../agent/form/tokenizer-form';
-import { getOperatorType } from './utils';
 
 type PipelineOperatorFormProps = {
   node: RAGFlowNodeType;

web/src/hooks/parser-config-utils.ts

@@ -4,6 +4,20 @@
  * and merge unknown fields into the `ext` field for flexible configuration.
  */
 
+/**
+ * Pipeline parser configs are keyed by operator id (e.g. "Parser:xxx"), so a
+ * top-level key containing ":" marks the pipeline structure, which must be
+ * sent as-is instead of being reshaped by extractParserConfigExt.
+ */
+export const isPipelineParserConfig = (
+  parserConfig: Record<string, any> | undefined,
+): boolean => {
+  if (!parserConfig || typeof parserConfig !== 'object') {
+    return false;
+  }
+  return Object.keys(parserConfig).some((key) => key.includes(':'));
+};
+
 /**
  * Extracts Raptor configuration with extra fields merged into ext.
  * @param raptorConfig - The raptor configuration object

web/src/hooks/use-document-request.ts

@@ -39,7 +39,10 @@ import {
   useGetPaginationWithRouter,
   useHandleSearchChange,
 } from './logic-hooks';
-import { extractParserConfigExt } from './parser-config-utils';
+import {
+  extractParserConfigExt,
+  isPipelineParserConfig,
+} from './parser-config-utils';
 import {
   useGetKnowledgeSearchParams,
   useSetPaginationParams,
@@ -502,6 +505,66 @@ export const useSetDocumentParser = () => {
   return { setDocumentParser: mutateAsync, data, loading };
 };
 
+/**
+ * Go-backend variant of useSetDocumentParser. The Go document endpoint takes
+ * `parser_id` (instead of the legacy `chunk_method`) and expects the
+ * pipeline-shaped parser_config (keyed by operator id) to be sent as-is.
+ * Keep it parallel to the Python version — the original hook stays untouched
+ * and can be dropped once the Python backend is retired.
+ */
+export const useSetDocumentPipelineParser = () => {
+  const queryClient = useQueryClient();
+
+  const {
+    data,
+    isPending: loading,
+    mutateAsync,
+  } = useMutation({
+    mutationKey: [DocumentApiAction.SetDocumentParser, 'pipeline'],
+    mutationFn: async ({
+      parserId,
+      pipelineId,
+      documentId,
+      datasetId,
+      parserConfig,
+    }: {
+      parserId: string;
+      pipelineId: string;
+      documentId: string;
+      datasetId: string;
+      parserConfig?: IChangeParserConfigRequestBody;
+    }) => {
+      const updateData: Record<string, unknown> = {
+        parser_id: parserId,
+        pipeline_id: pipelineId,
+      };
+     
+
+      if (parserConfig) {
+        updateData.parser_config = isPipelineParserConfig(parserConfig)
+          ? parserConfig
+          : extractParserConfigExt(parserConfig);
+      }
+
+      const { data } = await changeDocumentParser(
+        datasetId,
+        documentId,
+        updateData,
+      );
+      if (data.code === 0) {
+        queryClient.invalidateQueries({
+          queryKey: [DocumentApiAction.FetchDocumentList],
+        });
+
+        message.success(i18n.t('message.modified'));
+      }
+      return data.code;
+    },
+  });
+
+  return { setDocumentPipelineParser: mutateAsync, data, loading };
+};
+
 export const useSetDocumentMeta = () => {
   const queryClient = useQueryClient();
 

web/src/hooks/use-knowledge-request.ts

@@ -57,7 +57,10 @@ import {
   useGetPaginationWithRouter,
   useHandleSearchChange,
 } from './logic-hooks';
-import { extractParserConfigExt } from './parser-config-utils';
+import {
+  extractParserConfigExt,
+  isPipelineParserConfig,
+} from './parser-config-utils';
 import { useSetPaginationParams } from './route-hook';
 
 export const enum KnowledgeApiAction {
@@ -263,15 +266,6 @@ export const useDeleteKnowledge = () => {
   return { data, loading, deleteKnowledge: mutateAsync };
 };
 
-function isPipelineParserConfig(
-  parserConfig: Record<string, any> | undefined,
-): boolean {
-  if (!parserConfig || typeof parserConfig !== 'object') {
-    return false;
-  }
-  return Object.keys(parserConfig).some((key) => key.includes(':'));
-}
-
 export const useUpdateKnowledge = (shouldFetchList = false) => {
   const knowledgeBaseId = useKnowledgeBaseId();
   const queryClient = useQueryClient();

web/src/hooks/use-pipeline-operator.ts

@@ -0,0 +1,98 @@
+import { useFetchPipelineDslByPipelineId } from '@/hooks/use-agent-request';
+import { RAGFlowNodeType } from '@/interfaces/database/agent';
+import {
+  buildParserConfigFromNodes,
+  buildPipelineOperatorNodes,
+} from '@/utils/pipeline-operator';
+import { useEffect, useMemo, useRef, useState } from 'react';
+import { UseFormReturn } from 'react-hook-form';
+
+export const usePipelineOperatorNodes = (
+  pipelineId?: string,
+  pipelineParserConfig?: Record<string, any>,
+  isBuiltin = false,
+) => {
+  const { dsl, loading } = useFetchPipelineDslByPipelineId(
+    pipelineId,
+    isBuiltin,
+  );
+
+  const operatorNodes = useMemo(() => {
+    return buildPipelineOperatorNodes(dsl, pipelineParserConfig);
+  }, [dsl, pipelineParserConfig]);
+
+  return { operatorNodes, loading };
+};
+
+/**
+ * Resets parser_config when the selected pipeline changes, so stale configs
+ * from the previous pipeline are never submitted. Once the new pipeline's
+ * DSL has loaded, parser_config is seeded with the DSL defaults (i.e. what
+ * the operator tabs display).
+ *
+ * The very first pipeline seen is treated as the initial load: the form was
+ * already reset with the saved parser_config on mount, so it is left
+ * untouched.
+ */
+export const useResetParserConfigOnPipelineChange = (
+  form: UseFormReturn<any>,
+  pipelineId: string | undefined,
+  savedPipelineId: string | undefined,
+  operatorNodes: RAGFlowNodeType[],
+) => {
+  const previousPipelineIdRef = useRef<string>();
+
+  useEffect(() => {
+    if (!pipelineId) {
+      // Selection cleared (e.g. parse type switched): drop configs that
+      // belonged to the previously selected pipeline.
+      if (previousPipelineIdRef.current) {
+        previousPipelineIdRef.current = '';
+        form.setValue('parser_config', {});
+      }
+      return;
+    }
+
+    const isInitialLoad =
+      previousPipelineIdRef.current === undefined &&
+      pipelineId === savedPipelineId;
+    if (isInitialLoad || previousPipelineIdRef.current === pipelineId) {
+      previousPipelineIdRef.current = pipelineId;
+      return;
+    }
+
+    if (operatorNodes.length === 0) {
+      // The new pipeline's DSL is still loading — clear the previous
+      // pipeline's configs right away; defaults are seeded once it arrives.
+      form.setValue('parser_config', {});
+      return;
+    }
+
+    previousPipelineIdRef.current = pipelineId;
+    form.setValue('parser_config', buildParserConfigFromNodes(operatorNodes));
+  }, [form, pipelineId, savedPipelineId, operatorNodes]);
+};
+
+export const useActiveTab = (operatorNodes: RAGFlowNodeType[]) => {
+  const [activeTab, setActiveTab] = useState('');
+
+  useEffect(() => {
+    if (operatorNodes.length > 0) {
+      const firstTab =
+        (operatorNodes[0].data as Record<string, any>)?.operatorId ||
+        operatorNodes[0].data?.label ||
+        '';
+      const validTabs = operatorNodes.map(
+        (node) =>
+          (node.data as Record<string, any>)?.operatorId ||
+          node.data?.label ||
+          '',
+      );
+      setActiveTab((prev) => (validTabs.includes(prev) ? prev : firstTab));
+    } else {
+      setActiveTab('');
+    }
+  }, [operatorNodes]);
+
+  return { activeTab, setActiveTab };
+};

web/src/pages/dataset/dataset-setting/configuration/common-item.tsx

@@ -14,12 +14,9 @@ import {
   FormLabel,
   FormMessage,
 } from '@/components/ui/form';
-import { Radio } from '@/components/ui/radio';
 import { Spin } from '@/components/ui/spin';
 import { Switch } from '@/components/ui/switch';
-import { ParseType } from '@/constants/knowledge';
 import { useTranslate } from '@/hooks/common-hooks';
-import { useFetchBuiltinPipelines } from '@/hooks/use-agent-request';
 import { cn } from '@/lib/utils';
 import { history } from '@/utils/simple-history-util';
 import { t } from 'i18next';
@@ -30,7 +27,6 @@ import {
   FieldValues,
   useFormContext,
 } from 'react-hook-form';
-import { useTranslation } from 'react-i18next';
 import { useLocation } from 'react-router';
 import { DataSetContext } from '..';
 import { MetadataType } from '../../components/metedata/constant';
@@ -102,61 +98,6 @@ export function ChunkMethodItem(props: IProps) {
   );
 }
 
-export function BuiltinPipelineItem({
-  line = 2,
-  name = 'parser_id',
-}: {
-  line?: 1 | 2;
-  name?: string;
-}) {
-  const { t } = useTranslation();
-  const form = useFormContext();
-  const { options: builtinPipelineOptions } = useFetchBuiltinPipelines();
-
-  return (
-    <FormField
-      control={form.control}
-      name={name}
-      render={({ field }) => (
-        <FormItem className="items-center space-y-0">
-          <div
-            className={cn('flex', {
-              'items-center': line === 1,
-              'flex-col gap-1': line === 2,
-            })}
-          >
-            <FormLabel
-              required
-              className={cn('text-sm whitespace-wrap', {
-                'w-1/4': line === 1,
-              })}
-            >
-              {t('knowledgeConfiguration.builtIn')}
-            </FormLabel>
-            <div
-              className={cn('text-muted-foreground', { 'w-3/4': line === 1 })}
-            >
-              <FormControl>
-                <SelectWithSearch
-                  {...field}
-                  placeholder={t(
-                    'knowledgeConfiguration.chunkMethodPlaceholder',
-                  )}
-                  options={builtinPipelineOptions}
-                />
-              </FormControl>
-            </div>
-          </div>
-          <div className="flex pt-1">
-            <div className={line === 1 ? 'w-1/4' : ''}></div>
-            <FormMessage />
-          </div>
-        </FormItem>
-      )}
-    />
-  );
-}
-
 export const EmbeddingSelect = ({
   isEdit,
   field,
@@ -268,64 +209,6 @@ export function EmbeddingModelItem({
   );
 }
 
-export function ParseTypeItem({
-  line = 2,
-  name = 'parseType',
-}: {
-  line?: number;
-  name?: string;
-}) {
-  const { t } = useTranslate('knowledgeConfiguration');
-  const form = useFormContext();
-
-  return (
-    <FormField
-      control={form.control}
-      name={name}
-      render={({ field }) => (
-        <FormItem className="items-center space-y-0">
-          <div
-            className={cn('flex', {
-              'items-center': line === 1,
-              'flex-col gap-1': line === 2,
-            })}
-          >
-            <FormLabel
-              // tooltip={t('parseTypeTip')}
-              className={cn('text-sm  whitespace-wrap ', {
-                'w-1/4': line === 1,
-              })}
-            >
-              {t('parseType')}
-            </FormLabel>
-            <div
-              className={cn('text-muted-foreground', { 'w-3/4': line === 1 })}
-            >
-              <FormControl>
-                <Radio.Group {...field}>
-                  <div
-                    className={cn(
-                      'flex gap-2 justify-between text-muted-foreground',
-                      line === 1 ? 'w-1/2' : 'w-3/4',
-                    )}
-                  >
-                    <Radio value={ParseType.BuiltIn}>{t('builtIn')}</Radio>
-                    <Radio value={ParseType.Pipeline}>{t('manualSetup')}</Radio>
-                  </div>
-                </Radio.Group>
-              </FormControl>
-            </div>
-          </div>
-          <div className="flex pt-1">
-            <div className={line === 1 ? 'w-1/4' : ''}></div>
-            <FormMessage />
-          </div>
-        </FormItem>
-      )}
-    />
-  );
-}
-
 export function EnableAutoGenerateItem() {
   const { t } = useTranslate('knowledgeConfiguration');
   const form = useFormContext();

web/src/pages/dataset/dataset-setting/index.tsx

@@ -25,8 +25,9 @@ import ChunkMethodLearnMore from './chunk-method-learn-more';
 import LinkDataSource, {
   IDataSourceNodeProps,
 } from './components/link-data-source';
+import { ParseTypeItem } from '@/components/parse-type-form-field';
 import { MainContainer } from './configuration-form-container';
-import { ChunkMethodItem, ParseTypeItem } from './configuration/common-item';
+import { ChunkMethodItem } from './configuration/common-item';
 import { formSchema } from './form-schema';
 import { GeneralForm } from './general-form';
 import { useFetchKnowledgeConfigurationOnMount } from './hooks';

web/src/pages/dataset/dataset/dataset-table.tsx

@@ -14,6 +14,7 @@ import {
 import * as React from 'react';
 
 import { ChunkMethodDialog } from '@/components/chunk-method-dialog';
+import { DocumentPipelineDialog } from '@/components/document-pipeline-dialog';
 import { EmptyType } from '@/components/empty/constant';
 import Empty from '@/components/empty/empty';
 import { RenameDialog } from '@/components/rename-dialog';
@@ -28,6 +29,7 @@ import {
 } from '@/components/ui/table';
 import { UseRowSelectionType } from '@/hooks/logic-hooks/use-row-selection';
 import { useFetchDocumentList } from '@/hooks/use-document-request';
+import { isGoBackend } from '@/utils/backend-runtime';
 import { getExtension } from '@/utils/document-util';
 import { t } from 'i18next';
 import { pick } from 'lodash';
@@ -187,19 +189,29 @@ export function DatasetTable({
           ></RAGFlowPagination>
         </div>
       </div>
-      {changeParserVisible && (
-        <ChunkMethodDialog
-          documentId={changeParserRecord.id}
-          parserId={changeParserRecord.chunk_method}
-          pipelineId={changeParserRecord.pipeline_id}
-          parserConfig={changeParserRecord.parser_config}
-          documentExtension={getExtension(changeParserRecord.name)}
-          onOk={onChangeParserOk}
-          visible={changeParserVisible}
-          hideModal={hideChangeParserModal}
-          loading={changeParserLoading}
-        ></ChunkMethodDialog>
-      )}
+      {changeParserVisible &&
+        (isGoBackend() ? (
+          <DocumentPipelineDialog
+            parserId={changeParserRecord.chunk_method}
+            pipelineId={changeParserRecord.pipeline_id}
+            parserConfig={changeParserRecord.parser_config}
+            onOk={onChangeParserOk}
+            hideModal={hideChangeParserModal}
+            loading={changeParserLoading}
+          ></DocumentPipelineDialog>
+        ) : (
+          <ChunkMethodDialog
+            documentId={changeParserRecord.id}
+            parserId={changeParserRecord.chunk_method}
+            pipelineId={changeParserRecord.pipeline_id}
+            parserConfig={changeParserRecord.parser_config}
+            documentExtension={getExtension(changeParserRecord.name)}
+            onOk={onChangeParserOk}
+            visible={changeParserVisible}
+            hideModal={hideChangeParserModal}
+            loading={changeParserLoading}
+          ></ChunkMethodDialog>
+        ))}
 
       {renameVisible && (
         <RenameDialog

web/src/pages/dataset/dataset/use-change-document-parser.ts

@@ -1,11 +1,17 @@
 import { useSetModalState } from '@/hooks/common-hooks';
-import { useSetDocumentParser } from '@/hooks/use-document-request';
+import {
+  useSetDocumentParser,
+  useSetDocumentPipelineParser,
+} from '@/hooks/use-document-request';
 import { IDocumentInfo } from '@/interfaces/database/document';
 import { IChangeParserRequestBody } from '@/interfaces/request/document';
+import { isGoBackend } from '@/utils/backend-runtime';
 import { useCallback, useState } from 'react';
 
 export const useChangeDocumentParser = () => {
   const { setDocumentParser, loading } = useSetDocumentParser();
+  const { setDocumentPipelineParser, loading: pipelineParserLoading } =
+    useSetDocumentPipelineParser();
   const [record, setRecord] = useState<IDocumentInfo>({} as IDocumentInfo);
 
   const {
@@ -17,7 +23,12 @@ export const useChangeDocumentParser = () => {
   const onChangeParserOk = useCallback(
     async (parserConfigInfo: IChangeParserRequestBody) => {
       if (record?.id && record?.dataset_id) {
-        const ret = await setDocumentParser({
+        // The Go document endpoint takes `parser_id` and a pipeline-shaped
+        // parser_config; the Python one keeps the legacy payload shape.
+        const setParser = isGoBackend()
+          ? setDocumentPipelineParser
+          : setDocumentParser;
+        const ret = await setParser({
           parserId: parserConfigInfo.parser_id,
           pipelineId: parserConfigInfo.pipeline_id || '',
           documentId: record?.id,
@@ -29,7 +40,13 @@ export const useChangeDocumentParser = () => {
         }
       }
     },
-    [record?.id, record?.dataset_id, setDocumentParser, hideChangeParserModal],
+    [
+      record?.id,
+      record?.dataset_id,
+      setDocumentParser,
+      setDocumentPipelineParser,
+      hideChangeParserModal,
+    ],
   );
 
   const handleShowChangeParserModal = useCallback(
@@ -41,7 +58,7 @@ export const useChangeDocumentParser = () => {
   );
 
   return {
-    changeParserLoading: loading,
+    changeParserLoading: loading || pipelineParserLoading,
     onChangeParserOk,
     changeParserVisible,
     hideChangeParserModal,

web/src/pages/dataset/setting/hooks.ts

@@ -1,13 +1,16 @@
 import { ParseType } from '@/constants/knowledge';
-import { useFetchPipelineDslByPipelineId } from '@/hooks/use-agent-request';
 import {
   useFetchDatasetPipelineConfiguration,
   useUpdateKnowledge,
 } from '@/hooks/use-knowledge-request';
-import { RAGFlowNodeType } from '@/interfaces/database/agent';
 import { IConnector } from '@/interfaces/database/dataset';
 import { useDataSourceInfo } from '@/pages/user-setting/data-source/constant';
 import { checkEmbedding } from '@/services/knowledge-service';
+import {
+  getOperatorType,
+  transformApiConfigToForm,
+  transformFormConfigToApi,
+} from '@/utils/pipeline-operator';
 import { pick } from 'lodash';
 import {
   Dispatch,
@@ -15,20 +18,11 @@ import {
   useCallback,
   useEffect,
   useMemo,
-  useRef,
-  useState,
 } from 'react';
 import { UseFormReturn } from 'react-hook-form';
 import { useParams, useSearchParams } from 'react-router';
 import { z } from 'zod';
 import { formSchema } from './form-schema';
-import {
-  buildParserConfigFromNodes,
-  buildPipelineOperatorNodes,
-  getOperatorType,
-  transformApiConfigToForm,
-  transformFormConfigToApi,
-} from './utils';
 
 export function useHasParsedDocument(isEdit?: boolean) {
   const { data: knowledgeDetails } = useFetchDatasetPipelineConfiguration({
@@ -125,72 +119,6 @@ export const useFetchDatasetSettingOnMount = (
   return { knowledgeDetails, loading, sourceData };
 };
 
-export const usePipelineOperatorNodes = (
-  pipelineId?: string,
-  pipelineParserConfig?: Record<string, any>,
-  isBuiltin = false,
-) => {
-  const { dsl, loading } = useFetchPipelineDslByPipelineId(
-    pipelineId,
-    isBuiltin,
-  );
-
-  const operatorNodes = useMemo(() => {
-    return buildPipelineOperatorNodes(dsl, pipelineParserConfig);
-  }, [dsl, pipelineParserConfig]);
-
-  return { operatorNodes, loading };
-};
-
-/**
- * Resets parser_config when the selected pipeline changes, so stale configs
- * from the previous pipeline are never submitted. Once the new pipeline's
- * DSL has loaded, parser_config is seeded with the DSL defaults (i.e. what
- * the operator tabs display).
- *
- * The very first pipeline seen is treated as the initial load: the form was
- * already reset with the saved parser_config in useFetchDatasetSettingOnMount,
- * so it is left untouched.
- */
-export const useResetParserConfigOnPipelineChange = (
-  form: UseFormReturn<z.infer<typeof formSchema>>,
-  pipelineId: string | undefined,
-  savedPipelineId: string | undefined,
-  operatorNodes: RAGFlowNodeType[],
-) => {
-  const previousPipelineIdRef = useRef<string>();
-
-  useEffect(() => {
-    if (!pipelineId) {
-      // Selection cleared (e.g. parse type switched): drop configs that
-      // belonged to the previously selected pipeline.
-      if (previousPipelineIdRef.current) {
-        previousPipelineIdRef.current = '';
-        form.setValue('parser_config', {});
-      }
-      return;
-    }
-
-    const isInitialLoad =
-      previousPipelineIdRef.current === undefined &&
-      pipelineId === savedPipelineId;
-    if (isInitialLoad || previousPipelineIdRef.current === pipelineId) {
-      previousPipelineIdRef.current = pipelineId;
-      return;
-    }
-
-    if (operatorNodes.length === 0) {
-      // The new pipeline's DSL is still loading — clear the previous
-      // pipeline's configs right away; defaults are seeded once it arrives.
-      form.setValue('parser_config', {});
-      return;
-    }
-
-    previousPipelineIdRef.current = pipelineId;
-    form.setValue('parser_config', buildParserConfigFromNodes(operatorNodes));
-  }, [form, pipelineId, savedPipelineId, operatorNodes]);
-};
-
 export const useSaveDatasetSetting = () => {
   const { saveKnowledgeConfiguration, loading } = useUpdateKnowledge();
 
@@ -226,30 +154,6 @@ export const useSaveDatasetSetting = () => {
   return { handleSave, loading };
 };
 
-export const useActiveTab = (operatorNodes: RAGFlowNodeType[]) => {
-  const [activeTab, setActiveTab] = useState('');
-
-  useEffect(() => {
-    if (operatorNodes.length > 0) {
-      const firstTab =
-        (operatorNodes[0].data as Record<string, any>)?.operatorId ||
-        operatorNodes[0].data?.label ||
-        '';
-      const validTabs = operatorNodes.map(
-        (node) =>
-          (node.data as Record<string, any>)?.operatorId ||
-          node.data?.label ||
-          '',
-      );
-      setActiveTab((prev) => (validTabs.includes(prev) ? prev : firstTab));
-    } else {
-      setActiveTab('');
-    }
-  }, [operatorNodes]);
-
-  return { activeTab, setActiveTab };
-};
-
 export const usePipelineDataList = (sourceData: any[] | undefined) => {
   const { dataSourceInfo } = useDataSourceInfo();
 

web/src/pages/dataset/setting/index.tsx

@@ -1,4 +1,7 @@
+import { BuiltinPipelineItem } from '@/components/builtin-pipeline-form-field';
 import { DataFlowSelect } from '@/components/data-pipeline-select';
+import { ParseTypeItem } from '@/components/parse-type-form-field';
+import PipelineOperatorTabs from '@/components/pipeline-operator-tabs';
 import { Button, ButtonLoading } from '@/components/ui/button';
 import {
   Card,
@@ -12,9 +15,10 @@ import { Form } from '@/components/ui/form';
 import { FormLayout } from '@/constants/form';
 import { ParseType } from '@/constants/knowledge';
 import {
-  BuiltinPipelineItem,
-  ParseTypeItem,
-} from '@/pages/dataset/dataset-setting/configuration/common-item';
+  useActiveTab,
+  usePipelineOperatorNodes,
+  useResetParserConfigOnPipelineChange,
+} from '@/hooks/use-pipeline-operator';
 import { zodResolver } from '@hookform/resolvers/zod';
 import { useCallback, useEffect, useState } from 'react';
 import { useForm, useWatch } from 'react-hook-form';
@@ -26,15 +30,11 @@ import LinkDataSource, {
 import { formSchema } from './form-schema';
 import { GeneralForm } from './general-form';
 import {
-  useActiveTab,
   useConnectorHandlers,
   useFetchDatasetSettingOnMount,
   usePipelineDataList,
-  usePipelineOperatorNodes,
-  useResetParserConfigOnPipelineChange,
   useSaveDatasetSetting,
 } from './hooks';
-import PipelineOperatorTabs from './pipeline-operator-tabs';
 
 export default function DatasetSetting() {
   const { t } = useTranslation();

web/src/pages/datasets/dataset-creating-dialog.tsx

@@ -1,4 +1,6 @@
+import { BuiltinPipelineItem } from '@/components/builtin-pipeline-form-field';
 import { DataFlowSelect } from '@/components/data-pipeline-select';
+import { ParseTypeItem } from '@/components/parse-type-form-field';
 import { ButtonLoading } from '@/components/ui/button';
 import {
   Dialog,
@@ -28,10 +30,8 @@ import { useForm, useWatch } from 'react-hook-form';
 import { useTranslation } from 'react-i18next';
 import { z } from 'zod';
 import {
-  BuiltinPipelineItem,
   ChunkMethodItem,
   EmbeddingModelItem,
-  ParseTypeItem,
 } from '../dataset/dataset-setting/configuration/common-item';
 import { isGoBackend } from '@/utils/backend-runtime';